How Schema Markup Gives AI Something to Quote

A manufacturing client called me last spring convinced that schema markup was going to get his company quoted in ChatGPT. He had read that structured data was the key to AI search, and he wanted all of it, every type, on every page, by the end of the month. So I pulled his website. The schema markup he already had was injected by a tag manager script that never ran for any AI crawler. It was invisible to the exact systems he wanted to win.

That is the gap between what people believe about schema markup and what it does. The belief is selling a lot of work that does not pay off the way buyers think. So here is what schema markup is, how it works, and the one pathway that actually gives an AI system something to quote.

What schema markup is

Schema markup is structured data. It is code in the Schema.org vocabulary that you add to your website to label what your content means for search engines and AI. Search engines read schema markup to identify the entities, properties, and facts on a page, then use it to show rich results. Schema markup is not a direct input to AI answers, but it feeds the entity record AI systems quote from.

Everything past that is detail. The vocabulary comes from Schema.org, a shared standard the major search engines agreed on so there is one machine-readable way to say "this number is a price" and "this name is the author." You can write the code in three formats, microdata, RDFa, and JSON-LD, and Google recommends JSON-LD over microdata and RDFa because it is harder to break. Think of structured data as a translation layer. Your website is written for people. The schema markup is the same information, in code, written so search engines and AI can read it instead of having to interpret a page built for human eyes.

How schema markup works for search engines

A crawler hits your website, parses the HTML, and finds your JSON-LD block. Instead of guessing that "$299" is a price and "4.6 stars" is a rating, search engines read tags that say so outright. That explicit labeling is the entire mechanism. The code carries the same information your page already shows, so search engines understand the content faster and with less guesswork. It is the same job, done cleanly, that they were already attempting on the raw text.

What you get back is eligibility, not position. Schema markup makes a page eligible for rich results, also called rich snippets, the star ratings and prices and FAQ accordions that take up more room in search results. Rich snippets pull your information into the search results before anyone clicks. The labeling helps search engines understand your content and match it to more relevant queries, but the markup itself does not lift you up the page.

When Google reads the markup, it can present your content as a rich result in the search results: a recipe with a photo and a cook time, a product with a price and stars, an FAQ that expands. Google still decides whether to show the rich snippet, and it will only show one when the structured data matches the content a reader can see. So the markup gives Google and other search engines a cleaner way to understand your content, and the rich results are what that understanding looks like on the page.

Why schema markup is not a ranking factor

I raise this on most audits, because plenty of agencies still sell schema markup as a rankings lever. It is not one. Google's John Mueller has said it plainly: "Structured data won't make your site rank better." Structured data earns you display features and eligibility for rich results. It does not earn you a higher position in search.

The display still matters. 72.6% of top pages use schema markup, so on a competitive term you are not standing out with rich results, you are keeping pace. The website that skips schema markup is the one that looks thin next to nine competitors whose rich results show ratings, prices, and review counts in the search results.

The types of schema markup that earn rich results

Schema.org lists more than 800 types, but search engines generate rich results for only a short list, so you do not need most of them. The types of schema markup that earn their keep for a normal business are Organization, Local Business, Product, Article, Review, FAQ, and Event. Each type describes a different kind of content, and the rule is to use the most specific type that fits the page. Never mark up information the page does not actually show. Done right, these types of schema markup help search engines understand your content and make the page eligible for the rich results that fit it.

Never mark up information the page does not actually show.

Organization and Local Business schema markup

Organization schema tells search engines who your brand is: name, logo, contact information, and social profiles. You add Organization markup once for the whole website, and search engines use it to understand your brand as a single entity across every page. Local Business schema gives Google the address, hours, and phone it needs to place a physical location correctly in local search results and Maps. For a local business, this is the structured data that keeps your name, address, and phone consistent with what Google already has, which matters more than most owners realize.

Product, Review, and FAQ schema markup for rich results

Product schema pulls price, availability, and ratings into the search results, which is why ecommerce websites lean on it. Review schema adds the star rating and review count that make a listing stand out as a rich snippet. FAQ schema can surface questions and answers from your content directly under your result. When the type matches the page, the rich snippets are real: 2.7x more traffic from Recipe schema, and Eventbrite reported a 100% year-over-year jump in traffic from Google to its event listings after adding Event schema. None of that was a ranking boost. It was richer search results earning more of the clicks that were already there. Each of these types of schema markup gives search engines more of your content to show as rich results.

Rakuten (Recipe Schema)2.7 times

Eventbrite (Event Schema)100%

Source: searchenginejournal.com

What actually gives AI something to quote

Here is where the popular story falls apart. The assumption is that you add schema markup, an AI reads the structured data, and your facts land in the answer. Someone tested it. SEO consultant Mark Williams-Cook schema markup vs. visible content test. When he asked ChatGPT and Perplexity for the address, both returned the fake one.

That looks like proof schema works for AI. His conclusion was the reverse.

Why AI models read your schema markup as text

The models were not parsing the schema markup as structured data with rules. They were reading it as plain text, the same way they read everything else in the HTML. The JSON-LD format bought him nothing. The fabricated address was sitting in the page's code, and the model scraped those words like any other words. That is roughly how most large language models treat structured data. It gets stripped in training for many of them, and the middleware that feeds web pages to an LLM at answer time often strips it too.

There is one real exception. Dan Petrovic at DEJAN found that Gemini uses structured data through grounding, the step where Gemini checks Google's search index before it answers. Google's index parses structured data, so your schema markup reaches Gemini the long way around, through the index, not through the model reading your code directly.

The entity record behind your schema markup

So the pathway is indirect, and it runs through entities. Organization and Person schema feed Google's Knowledge Graph, the database of who and what your brand is. Google stores what it understands about your brand as entities there, and the cleaner the information you give it, the more accurately Google and the AI systems built on top of it understand and describe your business. When that entity record is clean, the systems that lean on it, entity cards, brand panels, and AI answers, describe you correctly. This is the same idea behind the semantic map your content is building: you are teaching the machine what you are, not asking it to rank you. Schema App reported a 19.72% AI Overview visibility lift. The thing an AI quotes is the fact in the entity record. Schema markup is one clean way to put it there.

Why AI crawlers never see your schema markup

Back to my manufacturing client. His schema markup was technically correct and completely wasted, because of how it loaded. The AI crawlers, GPTBot, ClaudeBot, and PerplexityBot, do not execute JavaScript. If your structured data is injected by Google Tag Manager or any client-side script, those crawlers fetch the raw HTML, find no JSON-LD, and move on. The markup only exists after a browser runs the script, and these crawlers never run it.

The fix is boring and it works. Put the schema markup in the page source as a static <script type="application/ld+json"> block, server-rendered, present before any JavaScript runs. I pulled the rendered HTML against the raw HTML for that client, showed him the markup vanishing from the raw fetch, and we moved it server-side. That is the kind of unglamorous problem technical SEO is actually made of.

It helps to remember what search engines are doing under the markup. A graph relations from text patent describes pulling entities and the relationships between them straight out of ordinary prose, no schema required. Google can build its understanding of your content from the words on your website alone. Schema markup is a shortcut that hands search engines the same graph cleanly, so they do not have to infer it. That is the honest way to think about structured data: not a secret channel into AI, but a faster path to a conclusion the engine was already trying to reach.

How to add schema markup to your website

The mechanics are easy. Generate the schema markup code with a tool or a CMS plugin, add it server-side into your website's HTML, and confirm search engines are reading it. You can add the code by hand or let a plugin add it for you, and most plugins output JSON-LD, the format Google prefers over older microdata and RDFa. If you run technical SEO for a real website, you already own every step.

How to add the right types of schema markup

Start with the types of schema markup that match what your pages actually are. An ecommerce website adds Product and Organization markup. A local business adds Local Business markup. A blog adds Article markup. Add structured data where it describes real content on the page, and skip it everywhere else. Google's own guidance for AI search is blunt: there is no special markup that gets you into AI features, and your structured data has to match the visible content. Do not add Review schema to a page with no reviews. Keep your name, address, and phone consistent with your Google Business Profile, because consistent information helps search engines understand which business you are, and the entity record is only as trustworthy as its least consistent source. The more accurately the content you add and your schema markup line up, the more types of rich results Google can show for the page.

How to test your schema markup

Never publish schema markup without testing it. Test the code through Google's Rich Results Test to see which rich results and rich snippets the page is eligible for, then test it once more through the Schema.org validator for a general syntax check. A single missing comma in the code is enough for search engines to reject the whole block, so test before you ship and test again after the page goes live. Watch the rich results reports in Google Search Console to confirm Google is reading your structured data and to catch errors before they spread across the website.

What schema markup can and cannot do

Schema markup earns rich results from Google, and those rich snippets earn clicks in the search results. It strengthens the entity record that AI systems quote from when they describe your brand. Both are worth doing, and on a competitive set of search results, doing them is table stakes rather than an edge.

What structured data will not do is act as a button that makes ChatGPT cite you. The models mostly read your website as text. The facts that survive into an AI answer are the ones that are clear in your content and consistent in your entity record, and schema markup is one tool for making them clear, not the whole job. That is the same lesson that runs through the Nashville SEO playbook: the durable wins come from search engines understanding what you are, and schema markup is how you spell it out without making them guess.

My client got his schema markup, server-side, on the pages that mattered, matched to what those pages actually said. He did not get quoted in ChatGPT the next week. He got an entity record Google could trust, which is the only version of this that was ever going to last.

By Michael McDougald